The recent publication of the preliminary results of the 1979 Census has

Transcription

1 The Economic and Social Review, Vol ll, No. 4, July 1980, pp The Use of the Irish Electoral Register for Population Estimation i i BRENDAN J. WHELAN AND GARY KEOGH* The Economic and Social Research Institute, Dublin Precis: A strong relationship is shown to exist between the number of registered electors in a county and the county's population in census years. This relationship can be used to estimate county populations in intercensal years. Total population estimates for the years 1961 to 1979 are derived and are used to construct a net migration series for the same period. This series is compared with migration series implicit in the Central Statistics Office intercensal population series. I INTRODUCTION The recent publication of the preliminary results of the 1979 Census has led to substantial revisions in the Central Statistics Office estimates of the population for the years The magnitude of these revisions, which in 1978 amounted to 90,000 persons or about 3 per cent of total population, has cast some doubt on the methodology used by the CSO to estimate the population in years when a census is not taken. Hughes (1980) has shown that the crucial problem in this methodology is the calculation of net annual migration, which the CSO estimates by reference to data on net passenger movement. The present paper attempts to derive an alternative method for the calculation of population estimates in intercensal years. The data used are the numbers of electors in each cojunty and county borough which are available annually. These figures have several desirable properties as a basis for population estimation: (a) They are available quite quickly, usually by May of the year in question. (b) They are independent of other data sources since a full count of the electorate is carried out each year. There is, therefore, no need to make assumptions about the pattern of net migration in deriving population estimates from the Electoral Register. Indeed, when the natural increase is subtracted from these estimates a series on net * Valuable comments on earlier drafts were received from B. M. Walsh, J. G. Hughes, J. Durkan, F. Kirwan and the referees. The authors are, of course, responsible for any errors that remain.

2 migration can be derived which is independent of the data on net passenger movement, which, in recpnt years, have proved unreliable as a basis for estimates of net migration. (c) The population estimates obtained can be broken down by county and county borough. This paper begins with!a description of the Electoral Register. It then goes on to discuss some ways in'which it can be used to make population estimates and to assess the quality of these estimates. Regional breakdowns are also given. The time series on net migration implicit in the population estimates is then examined and compared with that implicit in the population estimates published by the CSO. II THE ELECTORAL REGISTER I Each year the Franchise Sections of the various county councils and county borough councils publish a Register of Electors giving the names and addresses of those eligible' to vote in D iil, Local Government and European Assembly elections. This register comeis into effect on the 15th of April in each year. A count of the number of electors of each type in each Dail constituency and in each, county or county borough is available from the Franchise Section of the Department of the Environment by about May of the year in question. These counts are subsequently published in the Statistical Abstract. Appendix Table A.l shows the number of electors aged 21 and over in each county and county borough in the years This time period was selected because the number of elector;, in each county and county borough are not published for years prior to The boundaries of the boroughs have, in some cases, changed since 1961, but the newly defined boundaries were consistently used in both the censuses and the Electoral Registers. One aspect of these data proved troublesome. The voting age was lowered from 21 to 18 in 1973, which led tc a sudden increase in the recorded number of electors in that year. In order 1o maintain consistency, the data for the years 1973 to 1979 had to be adjusted. This was accomplished by estimating the number of persons aged in each year, using the 1971 Census combined with the Life Table (Statistical Bulletin, March 1976). These estimates were then subtracted from the Electoral Register as published to obtain the figures shown in Table A.l. Of course, this adjustment procedure assumes zero net migration in the year-old age group. It might be argued that this age group is one of those most prone to emigration and hence that it would be preferable to

3 assume some positive level of net emigration in making the adjustment. However, until the age distribution for 1979 is published, it cannot be established that net emigration did actually occur in this age group. Furthermore, it must be borne in mind that, in any given year, immigration at ages under 18 in previous years will serve to swell the younger cohorts, thus increasing the size of the age group in the year in question beyond what would be expected from the 1971 Census. Ill ESTIMATING THE POPULATION FROM THE ELECTORAL REGISTER Probably the simplest way of using the Electoral Register to estimate population is to assume that the ratio between Register and population is constant between censuses. Thus, to estimate the population in any year t, one calculates (P*/E*).. E t where E* denotes the number of persons on the Electoral Register, P* the population at the time of the previous census, and Et the number of persons on the Register in year t. The results of carrying out this exercise for each census year since 1946 are shown in Table 1. Table 1: Number of persons recorded in the Electoral Register (Ef) and the Census of Population (Pf) and total population as estimated from Ep (P^_.j/Et l) ' n c e n s u s years from 1951 to 1979 Number of persons on Census of the Population as estimated Year Electoral Register* Population from Electoral Register Difference ,823,864 2,955,107 _ ,805,711 2,960,593 2,925,695 34, ,762,097 2,898,264 2,889,085 9, ,691,084 2,818,341 2,781,463 36, ,726,221 2,884,002 2,876,900 7, ,780,796 2',978,248 2,975,181 3, ,011,811 3,364,881** 1 3,364, * The figures for the number of electors in 1979 has been adjusted as described above so as to reflect the likely number of electors aged 21 and over. ** Here and elsewhere in this paper, the figure used for the 1979 Census is that published in the Preliminary Report of the Census. This has recently been revised by 3,336, but this small revision does not materially affect any of our results. The difference between the ibensus figures and the estimate is over 30,000 in 1951 and 1961, but is less than 10,000 in 1956, 1966, 1971 and The two figures are almost identical in 1979, but it should be recalled that

4 ! 304 ECONOMIC AND S( >CIAL REVIEW i the Electoral Register shown excludes the estimated number of persons aged If there was substantial immigration of persons in this age group, then the figure for 1979 would be somewhat understated, so leading to a larger (positive) difference between census figure and estimate. It is noteworthy that the population was under-estimated in each of the years shown (i.e., the difference is always positive), suggesting that the ratio (P/E) has been rising over time., One explanation for this persistent tendency to underestimate might be that this ratio varies as between regions. Hence, differential rates of population change in the different regions might lead to persistent under-estimates. A disaggregated estimating method might, therefore, be more efficient than the simple ratio approach outlined above. Unfortunately, disaggregated (county) data for the Electoral Register are available only since The years in which both census data and the Electoral Register were available on a county basis were, therefore, 1961, 1966, 1971 and Given the problems posed by changes in the voting age, it was decided to omit 1979 from the data on which the estimating procedure was based. Initially we thought of deriving our estimating procedure from a regression equation of the following form: where i = county (i = 1...' 31), 30 Pit = a + bei t + 2J cjditj j=i t = time (t= 1961, l!966, 1971), Pit is the census population in county i in year t, Eit is the number oflocal government electors in county i in year t, a and b are constants to be estima ted, cj (j = ) are a set of regression coefficients, and Dj (j = ) are a set of dummy variables such that when i = j, Ditj = 1 and when i f j, Dj t j = 0 for all values of t. These dummy variables (Dj, j = 1, the intercept coefficient between counties. i, 30) were designed to test whether of the abo"e equation varied significantly as However, it soon became clear that heteroscedasticity was a problem, since the values of Pit and its variar.ee varied very substantially across counties. We, therefore, adopted the following specification:

5 30 Pit/Eit = a + _2 djditj where the dj(j = ) are a set of regression coefficients and the other symbols have the meanings assigned to them above. This equation embodies the basic assumption that the ratio of the population of a county to its Electoral Register is constant across all three census years. A number of factors may influence this ratio, including the age structure of the population (e.g., counties with above average percentages of persons aged 0-20 will have a higher than"average ratio) and the pattern of registration (e.g., persons living away from the family home may sometimes be included in the Electoral Register of their county of origin, rather than their county of residence). These and other deficiencies of the Electoral Register will not detract from its usefulness in estimating population, provided they remain relatively constant from year to year. Of course, if one were to use this regression equation to estimate population over long periods of time, the assumption of constancy in this ratio could be invalid since the age structure of a county might change substantially. However, these difficulties are unlikely to be serious when one is using the method as a means of estimating population in intercensal years. As soon as the results of a new census become available, the coefficients can be reestimated, thus incorporating the latest information on the age structure of each county in the estimating method. The intercept (a) and the coefficients (dj) were estimated using Stepwise Least Squares Regression, giving the equation shown in Table 2. Nine counties had insignificant coefficients and these are excluded from the equation. If one wishes to estimate the population of one of these excluded counties in a particular year, one simply multiplies its Electoral Register in that year by the constant, a = jl To obtain an estimate of the population of any of the counties included in Table 2 (i.e., those with significant coefficients) one adds the entry in Table 2 corresponding to the county to the constant and then multiplies the resulting figure by the Electoral Register. For instance, to estimate Mayo's population, one would multiply its Electoral Register by , (= ). The pattern shown in the coefficients is interesting. They are substantial and positive in Limerick County Borough and in Dublin County, about zero in the other county boroughs, 1 and substantial and negative in most of the western and north-western counties. Some explanations for this pattern are discussed below. We also tested a version of this equation which included additional dummy variables to : test for year-to-year variation, but none of these proved significant.

6 ECONOMIC AND SOCIAL REVIEW i Table 2: Estimates of dj in the prediction equation P/E = XdjDj based on data for 1961, 1966~and~1971~ ) County /county borough Coefficient t-value \ Limerick County Borough 0, Dublin County Kildare Cork County Borough i Kilkenny Wexford Louth Wicklow Limerick County -dl Tipperary North Riding Cork County i -(il Waterford County Mayo Kerry Monaghan -(i Longford Sligo Cavan i Clare Roscommon Donegal Leitrim Constant I R 2 = Overall F-value = with (22, 70) d.f. All coefficients are significant at the 5% level; the overall F-value and coefficients with t-statistics greater than 2.66 are significant at the 1% level. So far we have concentrated on estimates of the total population. An important feature of the present estimation method is the regional estimates which it provides. This represents an advantage over the CSO's annual intercensal estimates which are not broken down by county. Indeed, even in census years the regional figures only become available after some time, whereas the Electoral Register estimates are available by about May. Population estimates for the four Census years, broken down by county and county borough, are shown in Appendix Table A.2. together with the corresponding census figures. In general, the concordance between the two sets of figures is very good. For instance, the percentage errors in 1979 vary from 6.1 to 4.3.!

7 IV EVALUATION OF THE ESTIMATES Table 3 shows (a) the estimates derived from the ratio of Pt to Et, (b) those derived from the regression by aggregating across all 31 counties and county boroughs, (c) the CSO's pre-census estimates and (d) the CSO's postcensus estimates. One way of evaluating the accuracy of population estimates based on the Electoral Register is to ascertain if they differ from the CSO's final (post-census) estimate by lless than the revision the CSO makes between its own pre- and post-census estimates. On this criterion, it seems that, in the period , the CSO's estimates are in general more accurate than either of those based on the Electoral Register. The Root Mean Square Error* (RMSE) for the CSO's pre-census estimate in the inter-censal years is 10.3, compared with 21.5 for the Electoral Register ratio estimate and 24.0 for the regression estimate. Between 1972 and 1978, however, the situation is dramatically reversed; the RMSE for the ratio estimate falls to 19.5 and that of the regression to 16.3, while that of the CSO's pre-census estimates rises to However, in all periods the error of closure (i.e., the error in the year immediately preceding the census) is lower for the ratio estimate than for the CSO's pre-census estimate, j The regression estimate has a lower error of closure in each of the relevant years except Indeed, comparison between the estimates based on the Register and the CSO's post-census estimates may not be appropriate since the post-census estimate is likely to be based to some extent on the same data as the precensus figure. If, for instance, the annual short-term pattern of migration derived from the net passenger movement series was erroneous, both the pre- and post-census estimates would be affected in the same way. Thus the pre- and post-census estimates might agree, but fail to reflect the true population. There has been some speculation that the unprecedented growth in population between 1971 and 1979 was due in part to under-counting in the 1971 Census. Walsh (1979) has argued that this is unlikely in view of the regional pattern of the recorded increases in population. The Electoral Register data can be used to cast some light on this issue. The ratio estimate of the population for 1971 as shown in Table 3 is , about three thousand lower than the census figure. If substantial undercounting had occurred in 1971, one would expect this estimate to be considerably in excess of the census figure. Two slightly more elaborate tests were also carried out. In the first, the regression equation was re-estimated on the n * RMSE = ( 2 (X - Y) 2 lnf 2 where X,is the estimate being evaluated, Y the CSO's post-census i=l estimate and n the number of intercensal years in question.

8 Table 3: Total population (a) as estimated from (P*/E*)E t, (b) from the regression in Table 1, (c) the CSO's pre-census estimate, and (d) the CSO's post-census estimate (census years underlined) Year Estimate derived from Electoral Register by ratio (a) Estimate derived from Electoral Register by regression (b) CSO pre-census estimate (c) CSO post-census estimate (d) (e) = (d) - (a) Differences (f) = (d) - (b) (g) = (d)-(c) ~ "~~ " "" G 2913.G I2.G 15.u QH ,0 2926,0 7, , , Note: In columns (c) to (g) the figures for the intercensal years have been rounded to the nearest thousand.

9 basis of data for 1961, 1966 and 1979 and used to estimate The second used the data for 196: and 1966 to estimate The first test yielded an estimate for 1971 of thousand and the second thousand, both of which are slightly below the census figure. Thus, the present data lend no support to the contention that the 1971 Census seriously under-estimated the country's population. V POSSIBLE EXPLANATIONS FOR THE INTER-COUNTY VARIATIONS IN P/E It was noted above that dummy variables representing 22 of the 31 counties and county boroughs were significant in the regression. This raises the question of why there were systematic variations in the ratio as between the different counties. Two explanations suggested themselves: (a) that the variation was due to the age structure of the counties, specifically the proportion of the population unde'r 21, and (b) that counties might vary in the extent to which all those eligible to vote there were recorded in the census. For example, it seemed possible that in counties with high levels of emigration, some persons might be registered to vote in their county of origin, but were actually resident in, say, DJublin or some other city or town. The pattern of coefficients shown in Table 2 is consistent with both of these explanations. The positive coefficients in the county boroughs imply a higher population than would be expected on the basis of the Register. These areas have high rates of net immigration and high proportions of the population under 21. Negative coefficients are obtained in the western and north-western counties where net emigration is prevalent and where low proportions of the population are under 21. To examine the relative importance of the two explanations, two additional regressions were run and these are shown in Table 4, together with the original equation from Table 1. The first equation (i) regresses the ratio of total census population to! census population over 21 on the county dummies. This equation reflects the "pure" age structure as distinct from fluctuations due to the registration pattern. The second equation (ii) regresses the ratio of the population ovter 21 to the Electoral Register. This should eliminate the age structure effect and thus clarify the "pure" influence of the registration pattern. In both equations, only those variables are retained for which the coefficients are significant. In view of the higher R 2 's and larger number of coefficients in Equation (i), it seems that most of the variation which we set out to explain is due to variations in the age structure across counties. However, there appear to be quite sizeable variations in the boroughs of Water-

10 Table 4: Regressions of (i) P/P 21+, (ii) P 2 l+ /E, and (Hi) P/E on various county dummies for the years 1961, 19$6 and 1971 Regression (Hi) ^Regression (i) Regression (ii) (from Table 1) County/county borough P/P 21+ =a+ebjdj P /E=a+2bjDj P/E=a+2bjDj, j j J Limerick County Borough Coefficient t-value Coefficient t-value Coefficient t-value! Dublin County Kildare N.S Cork County Borough i. N.S. : Kilkenny N.S. N.S Wexford N.S. N.S Louth, N.S Wicklow N.S. N.S Limerick County N.S. N.S Tipperary North Riding N.S. N.S Cork County -' N.S Waterford County -] N.S Mayo N.S. N.S Kerry N.S Monaghan N.S Longford Sligo N.S Cavan N.S Clare N.S Roscommon N.S Donegal Leitrim Meath N.S. N.S. Offaly N.S. N.S. Carlo w N.S. N.S. Westmeath N.S. Waterford County Borough I N.S N.S. Dublin County Borough N.S N.S. Laois 1 N.S N.S. Gal way i N.S. N.S. N.S. Tipperary South Riding N.S. N.S. N.S. Constant R 2 ' Overall F-value P = Total population according to the Census of the Population. P = Total population aged 21 and over. E = Number of persons on the Electoral Register. N.S. means the coefficient hasja t-value insignificant at the 5% level.

11 ford, Cork and Limerick and in counties Dublin and Donegal which cannot be explained by the age structure of these counties alone. VI THE IMPLICIT NET MIGRATION SERIES Given any annual series of population estimates, one can derive a series on net migration, NM, from the identity P t = P t-1 + N I t-! + N M t-l where Pt is the population at the beginning of the year t and NIt is the natural increase in year t. The net migration series implicit in the Electoral Register estimates (EMR and EMG) and the CSO's estimates (CM) are shown in Table 5 and in Figure 1 (see p. 317). Table 5: Net migration as estimated from (a) ratio (i.e., (P*/E*) Ef) (b) regression in Table 1 and (c) the CSO's population estimates. Net migration implicit in Year* Electoral Register ratio estimate EMR Electoral regression EMG Register estimate CSO population estimates CM ** 31.1** 17.1** * April of the stated year to April of the next year. ** Estimates, assuming that births and deaths in the first quarter of 1979 equal those in first quarter of 1978.

12 Both the Electoral Register estimates exhibit a very similar pattern (r = 0.997). They are both considerably ;more variable than that implicit in the CSO's figures. It also appears that the Electoral Register estimates lag behind the CSO's figures by about two years. This is confirmed by computing the correlation between the two serie:> for various lags, as shown in Table 6. Table 6: Correlation coefficients between estimates of net migration based on (a) Electoral Register''and (b) CSO esvimdtes, with various lags. Variables! 0 1 t = lag (years) CM, EMR_ t 0.578* 1 CM, EMG_ t 0.599* C.604* C.60&* 0.646** 0.658** 0.634* 0.644* *Value significant at the 5 per cent level. **Value significant at the 1 per cent level. A lag of this type in the relationship between the two series is to be expected since it will probably take some time for a person to get on to the Electoral Register, whereas the CSO's estimates are based on net passenger movement data which are collected at the time of migration. However, the length of the lag appears greater than one wo aid have expected on a priori grounds. There seems to be a certain periodicity in EMR and EMG which does not occur in CM. The apparent length of the period (about 3-4 years) would lead one to suspect that the occurrence of elections, which are held about every four years, influences the accuracy of the Electoral Register. It would seem plausible that in years when an election takes place the Register is thoroughly checked and that it is, therefore, more accurate in the years immediately succeeding an election thm in other years. However, the type of' effect to be expected is not clear. On the one hand, the expectation of an election in a given year might induce more new voters to register than would otherwise do so. On the other hand, the checking of the register by the various political parties: prior to an election probably leads to a net decrease in thd number of names on the Register since those who have died and those who have moved out of the constituency will be eliminated. The timing of elections will also determine the year in which the effect manifests itself. For instance, if an election is held in February or March of a given year, jit is unlikely to affect the Electoral Register until April of the next year, whereas an election held in October probably influences the Register published in the following April. To give some idea o'f the effect of elections we have marked on the graph the dates on which general elections were held (i.e., October 1961, April 1965, June 1969, February 1973 and June 1977). Although the

13 pattern is not entirely uniform, troughs in the graph appear in general to follow elections. This would suggest that the elimination of ineligible voters is the stronger of the two effects described above. The high variability in the EMR and EMG series, as well as the observed lag, illustrate some of the problems involved in using the Electoral Register for estimating year-to-year changes in population. For instance, the method suggested here might not be sensitive enough to pick up the first year of a down-turn in population. However, it seems much more likely to be able to identify a reversal which lasts two or three years. Thus, the method would still have something to recommend it, since the alternative "passenger balance" method failed to identify just such a sustained reversal of trend in the years The variability in the migration estimates could perhaps be moderated by deriving them from a moving average of the intercensal population estimates, rather than from the annual figures. VII CONCLUSION We have shown that the Electoral Register can provide quite accurate estimates of total population, as well as some information about net migration. These estimates are available quickly and on a disaggregated (county) basis. For some purposes, an even more disaggregated estimate would be useful. In theory, it should be possible to carry out a regression analysis of the type outlined above on the basis of the 3,000 or so District Electoral Divisions (DEDs). However, there is a practical problem in carrying out the analysis, since the Electoral Register figures are not published on a DED basis. In fact, the polling districts or wards into which the Register is divided do not, in general, correspond to the DEDs. As part of the development of the RANSAM, the ESRI's computer-based sample selection system, work is proceeding to solve this problem by disaggregating each polling district into its constituent parts and re-combining these into DEDs. It is hoped to publish population estimates by DED in the near future. Of course, it is to be expected that the errors involved in estimating population might be greater at DED level than at county or national level. REFERENCES HUGHES, J. G., "What Went Wrong with Ireland's Recent Postcensal Population Estimates?", The Economic and Social Review, Vol. 11, No. 2, pp WALSH, B. M., "Expectations, Information and Human Migration: Specifying an Econometric Model of Irish Migration to Britain", Journal of Regional Science, Vol. 14, No. 1, pp WALSH, B. M., "Some Implications of Preliminary Results of the Census of Population 1979", Lecture to the Irish Planning Institute, October, 1979.

14 Table Al: Numbers of local government electors in each county and county borough, County boroughs n. n -, «Cork 45,783 45,782 45,424 45,940 46,148 Dublin 312, , , , ,538 Limerick 27,328 27,921 28,058 29,427 29,732 Waterford 16,387 16,652 16,807 16,968 17,236 68, ,363 30,188 17,389 69, ,305 30,204 17,628 70, ,103 30,348 17,712 70, ,583 30,332 17,918 Counties -Garlow - Cavan Clare Cork Donegal Dublin Galway j Kildare Kilkenny Laoighis Leitrim Limerick Longford Louth Mayo Meath Monaghan Offaly Roscommon Sligo Tipperary (N.R.) Tipperary (S.R.) Waterford Westmeath Wexford Wicklow 19,017 36,122 47, ,369 72, ,767 86, ,524 36,444 26,773 22,754 50,346 18,986 41,729 77,068 37,920 29,470 30,111 38,279 34,557 32,603 41,353 27,518 30,265 49,570 36, ,786-18,753 19,158 35,738 35,119 34,827 46,788 46,840 47, , , ,298 72,378 71,793 72, , , ,934 86,593 86,328 86, nc\ Q A K. 7 1 K no ^198 37i007 56,547 36,0/4 36,494 26,446 26,041 26,506 22,277 21,786 21,340 49,915 49,664 49,472 19,929 19,527 18,477 42,443 42,326 41,374 75,376 75,481 72,685 37,918 47,994 38,512 29,293 28,857 28,528 29,889 29,932 29,906 37,816 35,673 37,238 34,065 33,379 33,241 32,400 32,140 32,561 40,925 40,517 41,232 27,286 27,012 27,073 30,062 29,875 30,838 49,556 49,269 49,509 35,896.35,753 35,895-19^ ,558 47, ,868 71, ,794 88,941 *7 1 o *?» 1,UU4 37,196 3b,504 26,537 21,124 49,136 18,306 41,756 73,082 39,099 29,598 29,757 37,124 33,121 32,706 40,922 27,101 30,955 49,921 35,966-19,514-34,497 48, ,116 72, ,147 88,451 *7 I AAA 1 J.,-T-TT 37,812 36,424 26,335 21,102 49,808 18,579 41,785 72,572 40,023 29,230 29,998 36,933 32,540 33,897 40,710 27,162 30,448 50,332 36,852 I ,460 48, ,250 72, ,374 87,877 7u,-* 1** 38,345 36,414 26,388 20,704 49,644 18,424 41,969 71,366 40,237 29,284 30,035 36,528 32,373 33,116 40,728 27,594 30,580 50,534 37,464 ~ " 34,166 48, ,428 71, ,146 87,094 70,471 38,393 36,221 26,283 20,433 50,378 18,322 42,450 70,730 40,727 29,027 29,869 35,909 32,147 32,887 40,835 27,328 30,492 51,067 37,939 19,779 34,153 48, ,185 71, ,063 88,123 71,086 39,125 36,660 26,442 20,015 50,618 18,225 43,150 70,224 41,090 29,069 29,949 35,654 32,205 33,216 41,162 27,606 30,704 51,783 38,427

15 * 1974* 1975* 1976* 1977* 1978* 1979* County boroughs Cork 71,597 72,298 73,375 74,804 75,299 75,520 75,858 76,596 77,061 77,589 Dublin 329, , , , , , , , , ,931 Limerick 30,625 31,005 31,727 32,220 33,242 33,162 33,463 33,769 33,932 34,372 Waterford 18,152 18,290 18,520 18,670 18,780 18,598 18,386 18,311 18,319 18,444 Counties Carlow 19,878 20,025 20,289 20,367 20,830 21,218 21,374 21,952 22,421 22,757 Cavan 34,087 34,112 34,104 34,149 34,451 34,607 34,719 34,700 34,737 35,047 Clare 48,640 48,977 49,449 50,422 51,436 51,694 52,520 51,921 52,489 53,819 Cork 138, , , , , , , , , ,045 Donegal 71,890 72,167 72,597 71,631 71,475 71,768 72,640 73,936 74,986 76,328 Dublin 147, , , , , , , , , ,330 Galway 88,421 88,583 89,289 91,606 94,241 94,816 96,968 99, , ,819 Kerry 7-l-,-l-l-7 71, t8-]-7 7-2r ,267 73,980 75,074 76,388 76,688 77,930 Kildare 39,853 41,260 43,097 45,815 47,410 48,212 49,961 51,986 52,696 55,078 Kilkenny 36,620 36,845 36,813 37,065 37,641 38,146 38,718 39,492 39,900 40,991 Laoighis 26,309 26,233 26,359 26,335 26,894 26,698 26,898 27,592 28,212 25,098 Leitrim 19,731 19,638 19,421 18,872 19,029 18,710 18,699 18,967 18,997 19,122 Limerick 50,905 51,435 51,585 53,679 54,279 55,284 56,381 56,921 58,594 58,844 Longford 18,181 18,348 18,678 18,390 18,438 18,118 18,197 18,375 18,617 19,073 Louth 43,891 44,630 45,171 45,284 46,090 46,726 47,864 49,218 50,058 50,783 Mayo 69,314 68,917 68,961 69,023 69,938 70,207 71,939 73,528 73,895 75,250 Meath 41,772 42,406 43,509 44,320 46,258 47,383 48,760 50,283 51, Monaghan 29,092 29,397 29,588 29,014 29,006 29,667 30,442 30,653 30,870 31,401 Offaly 29,918 30,171 30,309 29,888 30,225 30,371 30,986 31,858 32,439 33,178 Roscommon 35,207 34,918 35,631 34,158 33,828 33,766 33,506 33,875 33,607 33,908 Sligo 31,821 31,080 32,238 31,611 32,289 32,428 32,491 33,268 33,563 34,421 Tipperary (N.R.) 33,194 33,142 33,225 33,560 33,695 33,403 33,843 34,670 34,920 35,317 Tipperary (S.R.) 40,981 41,034 41,345 41,581 41,776 42,106 42,624 43,436 43,783 45,131 Waterford 27,865 27,984 28,454 28,972 29,626 30,118 30,402 31,418 31,972 32,885 Westmeath 30,840 30,964 31,313 31,379 31,813 32,141 32,272 33,172 33,690 35,109 Wexford 51,661 51,793 53,098 53,269 53,856 54,116 54,941 55,548 56,370 57,658 Wicklow 38,664 39,204 40,252 41,344 42,795 44,307 45,352 47,263 47,492 49,776 *As explained in the text, these figures have been adjusted to take account of the lowering of the registration age from 21 to 18 years in 1973.

16 Table A.2*. Numbers of persons in each county and county borough in 1961, 1971 and 1979 (a) as recorded in the Census and (b) as estimated from the Electoral Register Census Estimate 1966 Census Estimate 1971 Census Estimate 1979 Census Estimate % error 1979 County boroughs Cork 77,980 80, , , , , , , Dublin 537, , , , , , , , Limerick 50,786 50,594 55,912 55,889 57,161 57,402 60,769 63, Waterford 28,216 28,102 29,842 29,820 31,968 31,365 32,617 31, Counties Carlow 33,342 32,612 33,593 33,464 34,237_? ,649 39,025.. _ C avail 56,594 56,293 54,022 53,761 52,618 53,161 53,706 54, Clare 73,702 72,937 73,597 74,220 75,008 75,164 84,823 82, Cork 252, , , , , , , , Donegal 113, , , , , , , , Dublin 180, , , , , , , , Galway 149, , , , , , , , Kerry 116, , , , , , , , Kildare fifi QK.KfiX Kilkenny 61,668 60,989 60,463 60,956 61,473 61,660 69,115 68, An_nco A.A. SO.C Afi. /1RO HO nm Lei trim 33,470 33,098 30,572 30,695 28,360 28,566 27,827 27, Limerick 82,553 82,137 81,445 81,260 83,298 83,914 96,605 96, Longford 30,643 29,833 28,989 29,194 28,250 28,831 30,777 29, Louth 67,378 68,961 69,519 69,053 74,951 73,755 86,180 83, Mayo 123, , , , , , , , Meath 65,122 65,028 67,323 68,634 71,729 72,721 90,589 91, Monaghan 47,088 46,517 45,732 46,139 46,242 46,402 50, Offaly 51,533 51,637 51,717 51,443 51,616 51,568 57,183 56, Roscommon 59,217 58,721 56,228 56, ,519 53,566 54,095 52, Sligo 53,561 54,207 51,263 51,043 50,275 49,895 54,609 53, Tipperary N.R. 53,596 52,979 53,843 55,082 54,337 53,955 58, Tipperary S.R. 70,126 70,915 68,969 69,813 69,228 70,368 75,215 77, Waterford 43,223 43,873 43,238 43,306 45,347 44,616 54,635 52, Westmeath 52,861 51,901 52,900 52,215 53,572 53,099 59,915 60, Wexford 83,308 82,709 83,437 83,980 86,351 86,418 96,259 96, Wicklow 58,473 59,593 60,428 60,801 66,295 64,682 83,793 82, Total 2,818,341 2,810,704 2,884,002 2,878,760 2,978,248 2,976,355 3,364,881 3,381,

17 00 I '